A method of generating accelerating potential energy and atomic force interpolations from on-the-fly quantum mechanical (QM) potential energy and atomic force data is presented. The interpolant (INT) has the properties of being local, adaptable, canonical, and completely general. Numerical studies examining the accuracy in the interpolated potential energy and atomic forces, total energy conservation, and acceleration have been performed on direct-dynamic QM simulations of diverse single molecules ranging in size from 3 to 26 atoms, including saturated and unsaturated hydrocarbon rings and chains with diverse functional groups. More extensive single-processor and parallel computational studies showed tremendous acceleration due to the interpolant in multilevel QM/MM simulations of mixed fuel systems of 100s to 1,000s of atoms where the simulations can be broken into evolving spatially-resolved groups. Load balance and number of core scaling studies were performed on 1,000 to 2,000 atom iso-butanol/O<sub>2</sub> systems of varying equivalency ratios and showed linear reduction in total CPU time and core saturation between 64 and 128 cores. These studies have been performed on the Accelerated MOLecular Dynamics with Chemistry (amoldc) program, designed to be a robust parallel multilevel QM/MM/INT simulation tool to study complex chemical reactions.

On-the-fly interpolations of quantum mechanical potentials and atomic forces incorporated into a massively parallel multilevel QM/MM simulation tool

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.